Starcloud Raises $250M More to Build Data Centers in Orbit

Starcloud has closed a $250 million extension to its March 2026 Series A, valuing the orbital data center startup at $2.3 billion. Manhattan West Ventures led the round, with Nvidia contributing $25 million and Cisco joining alongside Benchmark, EQT, Soma, NFX, 776, Cedar Capital, Goanna Capital, and Standard Capital (TechCrunch).

The extension arrives just over five months after Starcloud's initial $170 million Series A, which valued the company at $1.1 billion and was led by Benchmark (SpaceNews). That March round made Starcloud the fastest unicorn in Y Combinator history, crossing the $1 billion valuation threshold 17 months after its demo day (Y Combinator). The new capital is slated for a larger manufacturing facility and for advancing Starcloud-3, the company's largest orbital data center spacecraft, designed to fly on SpaceX's forthcoming Starship rocket (TechCrunch).

Starcloud, led by CEO Philip Johnston, is building satellites designed to run AI inference in orbit. Inference is the stage of AI where a trained model takes in new data and produces predictions or outputs, as opposed to training, where the model learns from data in the first place. Starcloud has achieved a notable technical milestone: it is the only company currently operating an Nvidia H100 data center GPU in orbit, and the first to train a model using it (TechCrunch). Its next commercial spacecraft, Starcloud-2, is scheduled for rideshare launches in 2027 into sun-synchronous orbit — a path that keeps satellites in continuous sunlight. Each satellite will provide 8 kW of compute capacity for orbital inference tasks (TechCrunch; SpaceNews). Customers for Starcloud-2's inference services include US government agencies (TechCrunch).

The company's ambitions extend well beyond a handful of satellites. Starcloud has filed a request with the FCC to operate 88,000 spacecraft (TechCrunch). Its website promotes the concept of space-based data centers offering 90% lower electricity costs and 24/7 solar-powered efficiency (Starcloud.com), framing orbital compute as a response to terrestrial data center energy demand (Y Combinator).

The launch landscape Starcloud must navigate is narrowing. SpaceX's Falcon 9 program is scheduled to end in 2028 as the company phases out the workhorse rocket and brings Starship online (TechCrunch). Blue Origin's New Glenn and ULA's Vulcan are not flying regularly, and Rocket Lab's Neutron has not yet reached the pad (TechCrunch). SpaceX CEO Elon Musk said the company will delay an attempt to catch a returning Starship rocket for a few months and will attempt to re-fly the vehicle for the first time at the end of 2026 or early 2027 (TechCrunch).

To hedge against launch constraints, Starcloud is considering buying a dedicated Falcon 9 launch and signing contracts with other launch providers to support future missions (TechCrunch). Starcloud-3's reliance on Starship adds a dependency on a vehicle that has not yet achieved operational reusability, let alone the launch cadence needed to support large-scale orbital infrastructure.

Nvidia's participation is notable beyond the dollar amount. The $25 million is modest relative to Nvidia's broader investment activity, but it signals that the dominant GPU vendor sees enough technical credibility in orbital inference to take a financial position. Starcloud's proven ability to operate an H100 in orbit and train a model on it gives that bet a concrete foundation rather than a purely speculative one.

For enterprise and government customers, the appeal of orbital compute rests on a proposition that has circulated in aerospace circles for years: continuous solar generation without atmospheric attenuation or nighttime cycles, combined with the ability to position compute physically close to satellite data sources, eliminating downlink latency for Earth-observation and sensor-heavy workloads. Starcloud's government customer base for Starcloud-2 suggests early demand is concentrated in exactly those use cases where the data never needs to come back to Earth in raw form.

The 88,000-spacecraft FCC filing is the clearest signal of Starcloud's long-term thesis: a constellation-scale orbital data center network. That filing will move through a regulatory process that has historically taken years for constellations a fraction of that size. Starcloud's nearer-term milestones, the 2027 Starcloud-2 launches, will be the first real test of whether orbital inference can serve paying customers at commercially viable economics.

The broader question is whether the unit economics of launching, powering, cooling, and maintaining GPU compute in orbit can approach terrestrial data center costs, even with the energy advantages Starcloud claims. The company has the funding, the silicon partnership, and a demonstrated orbital GPU. The next 18 months will reveal whether that combination is enough to turn orbital data centers from proof of concept into infrastructure.